The effect of environmental regulation on the locational choice of Japanese foreign direct investment

Abstract

This article assesses the impact of environmental regulation (ER) in host countries on Japanese foreign direct investment (FDI) decision-making. It tests the pollution haven hypothesis using data on national (ER) standards and Japanese inward FDI in five dirty industries (iron and steel industry, nonferrous metals industry, chemicals industry, paper and pulp industry, nonmetallic products industry). The results do not support the pollution hypothesis. On the contrary, inward Japanese FDI appears to be attracted to countries which have committed themselves to a transparent and stable environment regulatory environment, suggesting that the quality of the regulatory framework in terms of its certainty and transparency has a greater influence on foreign investors’ choice of location than the level of environmental regulatory measures.

Acknowledgements

We are grateful to the journal's referee and to Hulya Ulku for helpful comments on an earlier version of the article.

Notes

¹ Copeland and Taylor (2004) distinguish between the pollution haven effect and the pollution haven hypothesis. In the former case, a tightening of ERs will, at the margin, have an effect on trade and investment flows. In the latter case, the effect of cgqa dominates the influence of all other factors that affect trade and investment flows, and leads to a shift in pollution intensive industry from countries with more stringent regulations to countries with weaker ER.

² Although there has been a large number of empirical studies on the determinants of Japanese FDI (e.g. Co, 1997; In-Mee and Ozawa, 2001; Farrell et al ., 2004; Cassidy and O’Callaghan, 2005), very few have examined the relationship with ERs. An exception is the study by Friedman et al . (1992) who find that Japanese FDI in the US choose to locate in regions with relatively lax ERs. However, this study is restricted to Japanese FDI inflows to the US, and covers the earlier period 1977 to 1988.

³ The conditional logit model is appropriate when the data consist of choice-specific attributes. This model is widely used when three or more dependent variables are not consecutively ordered (McFadden, 1974; Green, 2000)

⁴ An alternative assumption would be that FDI first selects a region and then a country within the region. This would require the use of a nested logit model. We are grateful to the referee for drawing our attention to this point.

⁵ Participation in international environmental agreements is also used as a measure of environmental stringency in Smarzynska and Wei (2001) and Busse (2004).

⁶ National participation information for these five treaties is provided in World Bank (2000), World Development Indicators.

⁷ We accept the referee's comment that the stringency of environmental control may vary between different treaties. However, we do not have the information that would allow us to make this differentiation in the variable.

⁸ Dunning (1993) discusses these various factors and discusses the empirical evidence regarding their impact on FDI flows.

⁹ This also holds for studies of Japanese FDI (Chen, 1992; MITI, 1993, 1994; Mito, 1997; Economic Planning Agency, Japan,1993, 1994).

¹⁰ The main data used were taken from CASIO (2002) supplemented where necessary with data from the Japanese Vexillological Association (2002) and the Geographical Survey Institute of Japan (2002).

¹¹ Regional dummies were included in the regressions: these results can be provided on request.

¹² See Greene (2000) for details. McConnell and Schwab (1990) use the same approach in their empirical study.

¹³ For example, the Iron and Steel results suggest that, increasing the value of the ER Index from 9 to16, while holding all of the other parameters at their averages, would mean a 0.946% increase in the probability that a firm would choose to invest in the hypothetical average country.

¹⁴ For further details on the distinction between the categories in resource- and nonresource based industries, refer to UNIDO (1982). Van Beers and van den Bergh (1997) also make this distinction.

¹⁵ We tested for the nonlinearity of the relationship by adding ER squared as an additional variable. The results were less significant and confirmed the superiority of the linear specification.